cs.CL（2025-05-26）

📊 共 66 篇论文 | 🔗 12 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (51 🔗7) 支柱二：RL算法与架构 (RL & Architecture) (14 🔗5) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (51 篇)

#题目一句话要点标签🔗⭐
1 Thinking with Visual Abstract: Enhancing Multimodal Reasoning via Visual Abstraction 提出视觉抽象思维(VAT)方法，提升多模态大语言模型在视觉推理任务中的性能。 large language model multimodal chain-of-thought
2 SEMMA: A Semantic Aware Knowledge Graph Foundation Model 提出SEMMA以解决知识图谱推理中的语义不足问题 large language model foundation model
3 WebCoT: Enhancing Web Agent Reasoning by Reconstructing Chain-of-Thought in Reflection, Branching, and Rollback WebCoT：通过重构思维链提升Web Agent在反思、分支和回滚中的推理能力 large language model chain-of-thought
4 ALAS: Measuring Latent Speech-Text Alignment For Spoken Language Understanding In Multimodal LLMs 提出ALAS：一种用于评估多模态LLM中语音-文本潜在对齐的自动指标 large language model multimodal
5 MangaVQA and MangaLMM: A Benchmark and Specialized Model for Multimodal Manga Understanding 提出MangaVQA基准和MangaLMM模型，用于提升多模态漫画理解能力 multimodal
6 MedHELM: Holistic Evaluation of Large Language Models for Medical Tasks MedHELM：用于医学任务的大语言模型全面评估框架 large language model
7 Beyond Keywords: Evaluating Large Language Model Classification of Nuanced Ableism 评估大语言模型对细微歧视性语言的分类能力，揭示其在自闭症歧视识别上的局限性。 large language model
8 Multimodal Emotion Recognition in Conversations: A Survey of Methods, Trends, Challenges and Prospects 多模态对话情感识别综述：方法、趋势、挑战与展望 multimodal
9 Large Language Models for IT Automation Tasks: Are We There Yet? ITAB基准测试揭示大语言模型在IT自动化任务中，特别是Ansible脚本生成方面的局限性 large language model
10 WXImpactBench: A Disruptive Weather Impact Understanding Benchmark for Evaluating Large Language Models WXImpactBench：构建天气灾害影响理解基准，评估大语言模型在气候适应中的能力 large language model
11 THiNK: Can Large Language Models Think-aloud? THiNK：提出基于Bloom分类的多智能体反馈框架，评估并提升LLM的高阶思维能力。 large language model
12 Iterative Self-Incentivization Empowers Large Language Models as Agentic Searchers 提出EXSEARCH，通过迭代自激励提升大语言模型在知识密集型任务中的搜索能力 large language model
13 ResSVD: Residual Compensated SVD for Large Language Model Compression ResSVD：一种残差补偿的SVD大语言模型压缩方法 large language model
14 Language-Agnostic Suicidal Risk Detection Using Large Language Models 提出语言无关的自杀风险检测框架以解决现有方法局限性 large language model
15 Large Language Models Meet Knowledge Graphs for Question Answering: Synthesis and Opportunities 综述性研究：探讨大型语言模型与知识图谱融合在问答任务中的方法与机遇 large language model
16 MA-RAG: Multi-Agent Retrieval-Augmented Generation via Collaborative Chain-of-Thought Reasoning 提出MA-RAG多智能体框架，通过协同CoT推理解决复杂信息检索增强生成任务。 chain-of-thought
17 MiniLongBench: The Low-cost Long Context Understanding Benchmark for Large Language Models 提出MiniLongBench，一种低成本的长文本理解大语言模型评测基准 large language model ✅
18 FoodTaxo: Generating Food Taxonomies with Large Language Models FoodTaxo：利用大型语言模型自动生成食品分类体系 large language model
19 T^2Agent A Tool-augmented Multimodal Misinformation Detection Agent with Monte Carlo Tree Search 提出T^2Agent，一种基于蒙特卡洛树搜索的工具增强型多模态虚假信息检测Agent。 multimodal
20 Reasoning LLMs are Wandering Solution Explorers 揭示推理LLM缺乏系统性探索能力，指出其为游荡式问题解决者 large language model chain-of-thought
21 MemGuide: Intent-Driven Memory Selection for Goal-Oriented Multi-Session LLM Agents 提出MemGuide框架，通过意图驱动的记忆选择提升多轮对话LLM智能体的任务连贯性。 large language model chain-of-thought
22 OmniCharacter: Towards Immersive Role-Playing Agents with Seamless Speech-Language Personality Interaction OmniCharacter：提出一种无缝语音-语言个性化交互模型，实现沉浸式角色扮演Agent。 large language model ✅
23 SelfReflect: Can LLMs Communicate Their Internal Answer Distribution? 提出SelfReflect指标，评估LLM能否有效传达其内部答案分布的不确定性 large language model
24 Does quantization affect models' performance on long-context tasks? 系统评估量化对长文本LLM性能的影响，揭示任务、模型和量化方法的依赖性。 large language model
25 Paths Not Taken: Understanding and Mending the Multilingual Factual Recall Pipeline 揭示多语言LLM事实性召回pipeline，提出向量干预提升跨语言一致性 large language model
26 Gatsby Without the 'E': Crafting Lipograms with LLMs 利用大型语言模型生成有限制性文本：探索无'e'小说的创作 large language model
27 Amulet: Putting Complex Multi-Turn Conversations on the Stand with LLM Juries Amulet：利用LLM陪审团评估复杂多轮对话，提升评判准确性 large language model
28 HAMburger: Accelerating LLM Inference via Token Smashing HAMburger：通过Token压缩加速LLM推理，实现KV缓存和计算的亚线性增长。 large language model
29 Enhancing the Comprehensibility of Text Explanations via Unsupervised Concept Discovery 提出ECO-Concept框架，无需标注自动发现文本解释中的可理解概念。 large language model
30 FLAME-MoE: A Transparent End-to-End Research Platform for Mixture-of-Experts Language Models FLAME-MoE：开源混合专家语言模型研究平台，促进可复现性研究 large language model ✅
31 Estimating LLM Consistency: A User Baseline vs Surrogate Metrics 揭示LLM一致性度量与人类感知的偏差，提出logit集成方法提升对齐度 large language model
32 Improving the OOD Performance of Closed-Source LLMs on NLI Through Strategic Data Selection 通过策略性数据选择提升闭源LLM在NLI任务上的OOD泛化性能 large language model
33 Reasoning Is Not All You Need: Examining LLMs for Multi-Turn Mental Health Conversations 提出MedAgent框架与MHSD数据集，评估LLM在多轮心理健康对话中的表现 large language model
34 Pangu Light: Weight Re-Initialization for Pruning and Accelerating LLMs Pangu Light：通过权重重初始化加速和压缩大语言模型 large language model
35 UORA: Uniform Orthogonal Reinitialization Adaptation in Parameter-Efficient Fine-Tuning of Large Models UORA：大模型参数高效微调的均匀正交重初始化适配方法 large language model
36 Semantic-Preserving Adversarial Attacks on LLMs: An Adaptive Greedy Binary Search Approach 提出自适应贪婪二分搜索AGBS，用于LLM的语义保持对抗攻击。 large language model ✅
37 TrojanStego: Your Language Model Can Secretly Be A Steganographic Privacy Leaking Agent TrojanStego：提出一种基于语言模型隐写术的隐私泄露攻击方法 large language model
38 Named Entity Recognition in Historical Italian: The Case of Giacomo Leopardi's Zibaldone 针对历史意大利语，提出基于BERT和LLaMa的命名实体识别方法。 large language model
39 Multi-Domain Explainability of Preferences 提出多领域偏好可解释性方法，提升LLM对人类偏好的理解与对齐。 large language model
40 Inference-time Alignment in Continuous Space 提出SEA算法，通过连续空间梯度优化实现大语言模型推理时对齐。 large language model ✅
41 Grammars of Formal Uncertainty: When to Trust LLMs in Automated Reasoning Tasks 提出基于PCFG的LLM不确定性量化框架，提升自动推理任务可靠性 large language model
42 Training LLM-Based Agents with Synthetic Self-Reflected Trajectories and Partial Masking STeP：通过合成自反思轨迹和局部掩码训练LLM驱动的智能体 large language model
43 Emergent LLM behaviors are observationally equivalent to data leakage 大型语言模型涌现行为的解释：数据泄露而非社会规范 large language model
44 DeepDialogue: A Multi-Turn Emotionally-Rich Spoken Dialogue Dataset DeepDialogue：一个多轮、情感丰富的口语对话数据集，促进类人对话系统研究。 multimodal
45 APE: Selective Fine-tuning with Acceptance Criteria for Language Model Adaptation APE：基于接受准则的选择性微调方法用于语言模型自适应 large language model
46 Beyond Specialization: Benchmarking LLMs for Transliteration of Indian Languages 评估大型语言模型在印度语言音译中的表现 large language model
47 Improving Multilingual Math Reasoning for African Languages 针对非洲语言，研究者探索提升LLM在数学推理任务上的多语言能力的方法。 large language model
48 Deciphering Trajectory-Aided LLM Reasoning: An Optimization Perspective 从元学习视角解读：将LLM推理轨迹视为参数优化的伪梯度下降 large language model
49 Exploring Consciousness in LLMs: A Systematic Survey of Theories, Implementations, and Frontier Risks 系统性地探索LLM中的意识：理论、实现与前沿风险的综述 large language model ✅
50 MOLE: Metadata Extraction and Validation in Scientific Papers Using LLMs MOLE：利用大语言模型自动提取和验证科研论文元数据 large language model ✅
51 What Really Matters in Many-Shot Attacks? An Empirical Study of Long-Context Vulnerabilities in LLMs 研究揭示长文本情境下大语言模型在多示例攻击中的脆弱性，强调上下文长度是关键因素。 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (14 篇)

#题目一句话要点标签🔗⭐
52 Token-level Accept or Reject: A Micro Alignment Approach for Large Language Models 提出MARA：一种微调LLM的token级Accept-Reject对齐方法，降低计算成本。 preference learning RLHF DPO ✅
53 Enigmata: Scaling Logical Reasoning in Large Language Models with Synthetic Verifiable Puzzles Enigmata：通过合成可验证谜题提升大语言模型逻辑推理能力 reinforcement learning large language model
54 MT$^{3}$: Scaling MLLM-based Text Image Machine Translation via Multi-Task Reinforcement Learning 提出MT³框架，通过多任务强化学习提升MLLM在文本图像机器翻译任务上的性能。 reinforcement learning large language model multimodal
55 MaskSearch: A Universal Pre-Training Framework to Enhance Agentic Search Capability 提出MaskSearch预训练框架，提升LLM智能体在开放域多跳问答中的搜索能力 reinforcement learning curriculum learning distillation
56 Adaptive Deep Reasoning: Triggering Deep Thinking When Needed 提出自适应深度推理方法，根据问题复杂度动态切换长短推理链，提升LLM推理效率。 reinforcement learning large language model chain-of-thought
57 Incentivizing Strong Reasoning from Weak Supervision 提出弱监督激励方法，以低成本提升大语言模型的推理能力 reinforcement learning large language model chain-of-thought ✅
58 REARANK: Reasoning Re-ranking Agent via Reinforcement Learning 提出REARANK：基于强化学习的LLM推理重排序Agent，显著提升信息检索性能与可解释性 reinforcement learning large language model
59 Does Rationale Quality Matter? Enhancing Mental Disorder Detection via Selective Reasoning Distillation 通过选择性推理蒸馏提升心理健康检测效果，关注高质量理由 distillation large language model
60 R3-RAG: Learning Step-by-Step Reasoning and Retrieval for LLMs via Reinforcement Learning 提出R3-RAG，通过强化学习驱动LLM进行逐步推理和检索，提升RAG系统性能。 reinforcement learning large language model ✅
61 Enhancing Logical Reasoning in Language Models via Symbolically-Guided Monte Carlo Process Supervision 提出基于符号引导的蒙特卡洛过程监督方法，提升语言模型的逻辑推理能力 DPO direct preference optimization large language model
62 One-shot Entropy Minimization 提出单样本熵最小化方法，仅需少量数据即可显著提升大语言模型性能 reinforcement learning large language model ✅
63 Efficient Speech Translation through Model Compression and Knowledge Distillation 提出结合模型压缩与知识蒸馏的高效语音翻译方法，提升大模型部署效率。 distillation
64 Token Distillation: Attention-aware Input Embeddings For New Tokens 提出Token Distillation，通过注意力蒸馏为新token快速学习高质量嵌入表示。 distillation
65 REA-RL: Reflection-Aware Online Reinforcement Learning for Efficient Large Reasoning Models REA-RL：面向高效大型推理模型的反射感知在线强化学习 reinforcement learning ✅

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
66 Adaptive Classifier-Free Guidance via Dynamic Low-Confidence Masking 提出自适应无分类器引导（A-CFG），通过动态低置信度掩码提升生成模型的可控性。 classifier-free guidance

⬅️ 返回 cs.CL 首页 · 🏠 返回主页